Created on 2025-07-30 03:14
Published on 2025-07-31 02:38
As much as I can, I try to share grounded examples of using AI tools to do real world tasks. This week, an AI tool did something that both amazed and humbled me, and I want to share it.
As a part of managing a reimbursement program, we had a critical but tedious task to complete after we received cost reports from regions all over the state: compiling the data from nearly 300 individual spreadsheets into three master files. This was the first step to begin the process for program reimbursements. In the past, an entry-level team member would spend 3-4 days on this every year, opening, copying, and pasting data between spreadsheets. It was a classic manual data process, important for getting the process going but long, repetitive, and boring.
Putting my Python skills to use, I built a script to automate one program. With a little help from ChatGPT to expand the code for all the programs, I could consolidate all 300+ spreadsheets, run validation checks, and have perfectly organized data in about 30 seconds.
I had taken a multi-day task off my team's plate and created a faster, more reliable solution. I basically thought I perfected this specific workflow in an afternoon.
This week, I tried the new ChatGPT agent model. As a test, I uploaded six sample spreadsheets from my original test of the script I mentioned above and gave it a single-sentence prompt: "Consolidate these files, calculate the totals, and flag any records with funding percentages over 100%."
I expected to hit 'enter', leave to go heat up my dinner, and check on its progress. It was done before I even stood up.
Here’s a replay of the agent receiving the prompt and using python to accomplish the task.
The agent flawlessly:
Compiled all the names, amounts, and funding percentages.
Calculated all the sums and totals correctly.
Flagged the records with invalid data.
Provided a clean text summary of what it had done and the flagged records.
It was rewarding using my python skills to turn a tedious, multi-day task into a afternoon breeze, but it was actually disheartening to watch the ChatGPT agent take the files, a short prompt, and use python to make it almost effortless.
As someone who supervised teams doing this kind of work, this was a powerful moment. A huge portion of the routine financial monitoring (~20%) was based on these exact kinds of 'desk reviews' to check thresholds, validate totals, and ensure data reconciled.
This isn't about a script I wrote becoming obsolete. It's a preview of how drastically the nature of entry-level and analytical work is about to change. The barrier to executing complex data tasks is dropping to near zero.
#AIAutomation #ChatGPTAgent #AI #Automation #Python #WorkflowAutomation #AITools